topic Re: sample size calcs - why do these produce different results in Statistical Procedures

sample size calcs - why do these produce different results

msecic — Mon, 06 Apr 2015 21:16:50 GMT

data a;

input z p q e;

cards;

1.96 0.95 0.05 0.05

;

data b;

set a;

n=(z*z*p*q)/(e*e);

run;

proc print; run;

proc power;

onesamplefreq test=z method=normal

sides = 2

alpha = 0.05

nullproportion = 0.90

proportion = 0.95

ntotal = .

power = 0.9;

run;

Re: sample size calcs - why do these produce different results

lvm — Tue, 07 Apr 2015 20:47:32 GMT

Because your formula is not the one used by POWER. See the details at:

SAS/STAT(R) 9.3 User's Guide

Re: sample size calcs - why do these produce different results

msecic — Tue, 07 Apr 2015 20:55:31 GMT

right, I know, I understand that. But one is confidence interval estimation and the 'proc power' is hypothesis testing. They should be more in agreement, right?

Re: sample size calcs - why do these produce different results

lvm — Tue, 07 Apr 2015 20:58:45 GMT

Not necessarily. These are different things.

Re: sample size calcs - why do these produce different results

ballardw — Wed, 08 Apr 2015 16:02:26 GMT

POWER is involved with Type II error. The other approach is not.